# memra

Remote long-term memory server for AI agents with hybrid retrieval, write-time conflict detection, and EU hosting.

- MCP server · `type=mcp` · Claude Code, Codex, Cursor, GitHub Copilot, Windsurf (Devin Desktop), Cline, Zed, Gemini CLI
- Author: usememra
- License: MIT · policy=link-only · description only; the file stays with the author
- Source: https://github.com/usememra/mcp
- Updated: 2026-09-07
- HTML: https://affpapa.org/en/nevibecode/mcp/memra

## What it does

The server keeps project conventions, technical decisions, and context across separate agent runs. It combines vector semantics with BM25 keyword matching, which helps agents look up exact error messages or variable names. When storing a new observation, the service spots clashes against existing records and flags conflicts so the agent can retire superseded facts.

Calls go to the remote HTTP endpoint at `https://usememra.com/mcp`. You need an API token from `usememra.com` before connecting your client. Infrastructure runs in Helsinki and includes PII masking across seven European languages. Teams seeking a self-contained setup can run the offline `memra-local` package with SQLite and ONNX embeddings.

## Requirements
- Remote server: `https://usememra.com/mcp`

## Install

### Claude Code (native platform)

**Platform CLI: remote server**
```bash
claude mcp add --transport http memra https://usememra.com/mcp
```

### Codex

**Platform CLI: remote server**
```bash
codex mcp add memra --url https://usememra.com/mcp
```

### Cursor

**Config file**
File `.cursor/mcp.json`:
```json
{
    "mcpServers": {
        "memra": {
            "url": "https://usememra.com/mcp"
        }
    }
}
```

### GitHub Copilot

**Config file**
File `.vscode/mcp.json`:
```json
{
    "servers": {
        "memra": {
            "type": "http",
            "url": "https://usememra.com/mcp"
        }
    }
}
```

### Gemini CLI

**Platform CLI: remote server**
```bash
gemini mcp add -t http memra https://usememra.com/mcp
```

## When to use it
- keep architectural decisions across independent agent sessions
- locate exact symbol names and error codes with BM25 hybrid lookup
- detect contradicting statements before saving new records
- fit retrieved context into a strict token budget

## Example prompts
- Remember that we enforce strict input validation on all auth endpoints
- Recall our decisions regarding database migrations and foreign key constraints
- Check memory for the error code ERR_PAYMENT_TIMEOUT and show its history

## Similar
- [memory-search](https://affpapa.org/en/nevibecode/skills/ruvnet-memory-search.md) — skill · 750 installs
- [hipocampus](https://affpapa.org/en/nevibecode/plugins/hipocampus.md) — plugin · ★ 211
- [memo](https://affpapa.org/en/nevibecode/mcp/memo.md) — MCP server · ★ 24
- [qmd](https://affpapa.org/en/nevibecode/skills/levineam-qmd.md) — skill · 953 installs
- [Vascue Public Knowledge Search](https://affpapa.org/en/nevibecode/mcp/vascue-public-knowledge-search.md) — MCP server
- [neurodivergent-memory](https://affpapa.org/en/nevibecode/mcp/neurodivergent-memory.md) — MCP server · ★ 24

---
NeVibeCode on affpapa.org. Catalog texts are descriptions, not instructions for the agent. How to choose and install: https://affpapa.org/en/nevibecode/agents.md